Research On Cerebrovascular Skeleton Extraction Using Level Set Model

One of the major goals of computer vision and machine intelligence is the research of flexible and efficient methods for shape representation.An important approach for representing the structure of a shape is skeletonization.A skeleton is a compact representation of a shape that maintains its topology,it is of importance in a wide range of applications such as shape matching and retrieval,character animation and morphing,and medical image analysis.Because of the ambiguity,complexity and diversity of medical cerebrovascular image,the conventional skeleton detection algorithm extracted the centerlines which are not continuous,and brings the disadvantage of sensitive to boundary noise,less robust,and finally are of too much burr.Therefore,the research on how to extract the skeleton of object region of medical cerebrovascular images accurately is the starting point of this paper.The goal of this research is acquiring more accurate and rapid centerlines extraction algorithm for medical cerebrovascular images,which is help to increase the accuracy of 3D medical image reconstruction,identification of object region, and better to support medical research and clinical diagnosis.This dissertation made a widely review and survey of elementary knowledge of medical images,the current main methods and techniques of centerline extraction,both fully developed and under exploration,and analyzed the respective application spheres, advantages and disadvantages.Then studied the principle of the Level Set model in detail, and summed up the different characteristics of the different Level Set models.Combining with the inherent characteristics of medical gray images and color images,this paper proposed two new centerline extraction algorithms aiming at gray images and color images separately.In the method of skeleton extraction on gray cerebrovascular image,we have employed two different medial functions:the Euclidean distance field and a variant of the magnitude of the gradient vector flow(GVF),resulting in two different energy functions. The first energy controls the identification of the shape topological nodes from which curve skeletons start,while the second one controls the extraction of curve skeletons.It avoids locating and classifying skeleton junction points in order to guide the extraction of curve skeletons,and is completely automated because all its parameters are analytically derived.In the method of skeleton extraction on color cerebrovascular image,We use the HSV color space which is close to the human understand,re-designed a speed function based on color gradient function to color images to replace the traditional method of gray-scale gradient to stop the movement of the contour line,and making the noise caused by the color gradual change and weakening of the target marginal neglect combining regional color statistic characteristics.We use this new model to the color medical image skeleton extraction algorithm.Qualitative and quantitative analysis of the pros and cons of the novel model was given lastly.The last part of this paper generalized the work of my graduation topic,and put forward the further research ideas.